sti2025_pres

ERC as a Trendsetter?

Measuring Similarity in the Objectives of Funded Research in Europe

Radim Hladík,¹ Marc Bertin,² Chérifa Boukacem-Zeghmouri²

¹ Centre for Science, Technology, and Society Studies (CSTSS)
Institute of Philosophy, Czech Academy of Sciences
<hladik@flu.cas.cz>, <@hlageek@bsky.com>

² ELICO
Université Claude Bernard Lyon 1
<marc.bertin@univ-lyon1.f>,
<cherifa.boukacem-zeghmouri@univ-lyon1.f>

https://hlageek.github.io/reports/sti2025

Excellence in science

scientific excellence is as ubiquitous in science policy and research evaluation as it is challenging to define
often a “know it when you see it” concept
in ever-scarcer science funding, excellence - of proposals and/or applicants - is used as a fundamental criterion for funding decisions

How can we research excellence?

from philosophy of science to sociology of science
- even if excellence lacks objective definition, we can work with a practical definition
  - excellent science is what ERC think it is
- which sociologically translates to:
  - excellent science is what ERC funds

Does ERC set trends in the European Research Area?

why do we care about ERC’s definitions?
- ERC grants have high volume of economic and symbolic capital
  - projects of long duration and substantial amount
  - institutional impacts
    - universities adjust internal regulations to attract and carry-out ERC projects (Edler et al. 2014; Cruz-Castro, Benitez-Amado, and Sanz-Menéndez 2016)
    - funders develop ERC-inspired programs (e.g. Sweden, Czechia)
  - even when objective conditions (duration, amount) of projects are equivavalent, ERC holders are considered as elites (Hoenig 2017; Edlund 2024)

Does ERC set trends in the European Research Area?

If a practical definition of scientific excellence is emerging from the activities of ECR, we expect other funders to fund more ERC-like projects.

RQ1: Are research projects funded by national funders becoming more similar over time to transnational ERC projects?
RQ2: Does the epistemic breadth of science supported by national funders change along with the epistemic breadth of ERC-supported science?

Data

A convenience sample from European funding agencies providing project metadata including abstracts in English.

9 agencies / 8 countries + transnational ERC

European Research Council (ERC), European Union
Agence nationale de la recherche (ANR), France
Deutsche Forschungsgemeinschaft (DFG), Germany
Fonds Wetenschappelijk Onderzoek (FWO), Belgium
Grantová agentura České republiky (GAČR), Czech Republic
Norges Forskningsråd (NFR), Norway
Schweizerischer Nationalfonds / Fonds National Suisse (SNSF), Switzerland
UK Research and Innovation (UKRI), United Kingdom
Vetenskapsrådet (VR), Sweden

Analysis dataset

automated language checks
aligning years (some agencies publish year of the call, some date of the project start, or both)
project duration >= 3 years & <= 6 years
98455 project abstracts included in the analyses

Year	ANR	GACR	DFG	ERC	NFR	FWO	SNSF	UKRI	VR
2009	238	670	317	273	53	129	612	1646	995
2010	753	470	439	483	41	157	729	1587	894
2011	727	442	428	695	70	140	669	1353	872
2012	709	459	682	780	43	148	781	1808	1065
2013	596	404	1026	881	41	219	773	1920	1019
2014	761	338	1237	902	45	182	741	1666	821
2015	767	538	1280	935	58	192	778	1631	791
2016	953	577	1468	928	50	220	696	1600	739
2017	1059	732	1451	935	44	259	635	1499	634
2018	1129	585	1666	1001	22	329	530	1394	718
2019	1150	681	2237	918	22	330	556	1324	720
2020	1211	544	2552	908	27	346	488	1181	720
2021	1770	329	2688	973	106	328	585	1485	739
2022	1698	475	2419	1043	11	328	585	1583	643

Methods

descriptive
abstract embeddings obtained via SPECTER2 (2023 update)
- BERT-type language model of scientific texts trained over a citation graph
2D dimension reduction via UMAP
non-ERC projects assigned to the ERC panel of the closest ERC project
- validated by reassignment of ERC projects (M = 0.96, SD = 0.03, range [0.85, 1.00])
RQ1: measuring mean pairwise similarity of non-ERC projects to ERC projects within panel-year combinations
RQ2: measuring the spread of projects in the 2D embedding space via convex hull area

UMAP projection of projects

Dots signify average coordinates of ERC panels (true and estimated).

cf. maps of science - Rafols 2025; Hladík & Renisio 2025; Klavans & Boyack 2009

Density of projects in the UMAP projection

Trends in the similarity to ERC

Note: For ERC, within-similarity is shown.

Trends in the epistemic breadth

Preliminary results

quantitative evidence of growing similarity of nationally funded projects to ERC
- caveat: causality cannot be determined
  - is ERC setting the trend or is it following a confounder along with other funders?
the assimilation process is slow, i.e. a long-term perspective is important
the epistemic scope appears unrelated to the assimilation process
- rather a function of the number of projects funded (i.e. budget)

Discussion

comparative lense: research on research funding benefits from transcending single-funder or single-country approaches and taking heed of transversal patterns (Aagaard et al. 2021)
data on research funding: lack of metadata standards and data sharing
excellence
- an emerging norm for presenting research objectives to peers
- competitive research funding as a channel for dissemination of the norm
- possibly performative (in both senses - self-actualizing and without substance)
- stratification among PIs - high ERC prestige
policy: is the growing standardization of excellence a desirable policy goal?

Acknowledgements

https://hlageek.github.io/reports/sti2025

This work has been funded by a grant from the Programme Johannes Amos Comenius under the Ministry of Education, Youth and Sports of the Czech Republic, CZ.02.01.01/00/23_025/0008711.